10654
Langmuir 2003, 19, 10654-10659
Nanocrystal Surface Structure Analysis by Analytical Ultracentrifugation Helmut Co¨lfen,*,† Shay Tirosh,‡ and Arie Zaban‡ Max-Planck-Institute of Colloids and Interfaces, Colloid Chemistry, Research Campus Golm, Am Mu¨ hlenberg, 14476 Golm, Germany, and Chemistry Department, Bar Ilan University, Ramat-Gan, 52900, Israel Received April 25, 2003. In Final Form: October 8, 2003 Sedimentation velocity experiments on two TiO2 model colloids with similar properties but different surface structures were performed as a function of the solution pH in order to differentiate between similar size particles. The particle sedimentation velocity is highly sensitive to the surface structure, which is indicative of the exposed crystal face of the nanocrystals. Increase of the pH from 1 to 3 resulted in aggregation of all particles in one sample, whereas only partial aggregation occurred for the other, although the ζ-potential of both samples is almost identical in that pH range indicating particle stability. Although the particles are not distinguishable by the conventional methods for particle charge determination, they are clearly different in terms of their sedimentation coefficient distributions. Furthermore, analytical ultracentrifugation (AUC) reveals a dependence of the onset of large aggregate formation on the particle surface. This suggests that AUC has the potential to quantitatively determine differences in the particle surface structure even for polydisperse samples with constant average charge where conventional ζ-potential measurements yield only a constant average value. As ultracentrifugation yields distributions, we further discuss whether a combination with a second independent method like flow-field-flow fractionation can yield particle size and charge distributions in a global analysis approach.
Introduction The research, development, and use of nanosize materials have increased in recent years as part of the dramatic growth of nanotechnology. The literature provides numerous applications in which the special sizeinduced properties of nanoparticles are utilized.1-3 Both the perfection of existing applications and the design of new systems benefit from the ability to tune properties, such as the spectral response, energy levels, and electronic structure, via control of the material size.1-3 However, the most fundamental characteristic of nanoparticles is their high surface-to-volume ratio, which turns any application consisting of nanoparticles into a high surface area system. The exposed surfaces of crystalline particles depend on the crystal structure and the cleaving orientation.3-5 From bulk crystals, we know that the properties of the different surface structures vary significantly.3 Thus, two nanocrystals that are considered similar in terms of size, size distribution, shape, and crystal structure may perform considerably differently only because their surface structures are different.4-7 * To whom correspondence should be addressed. E-mail: coelfen@ mpikg-golm.mpg.de. † Max-Planck-Institute of Colloids and Interfaces. ‡ Bar Ilan University. (1) Nanoscale Materials in Chemistry; Klabunde, K. J., Ed.; John Wiley & Sons: New York, 2001. (2) Nanocrystalline Metals and Oxides, Selected Properties and Applications; Knauth, P., Schoonman, J., Eds.; Kluwer Academic: New York, 2001. (3) Hand Book of Nanostructured Materials and Nanotechnology; Nalwa, H. S., Ed.; Academic Press: San Diego, 2001. (4) Shklover, V.; Nazeeruddin, M. K.; Zakeeruddin, S. M.; Barbe, C.; Kay, A.; Haibach, T.; Steurer, W.; Hermann, R.; Nissen, H. U.; Gra¨tzel, M. Chem. Mater. 1997, 9, 430. (5) Oliver, P. M.; Watson, G. W.; Kelsey, E. T.; Parker, S. C. J. Mater. Chem. 1997, 7, 563. (6) Penn, R. L.; Banfield, J. F. Am. Mineral. 1998, 83, 1077. (7) Kato, K.; Torii, Y.; Taoda, H.; Kato, T.; Butsugan, Y.; Niihara, K. J. Mater. Sci. Lett. 1996, 15, 913-915.
For example, we found that the surface affects the stability and the photoelectrochemical activity of nanosize TiO2.8 The transformation temperature from anatase to the more stable structure, the rutile, was altered by more than 150 °C, and the performance of dye-sensitized solar cells consisting of these particles improved, only due to surface effects on the nanoparticles.8 In the above study, the surfaces of the nanoparticles were examined by dark field transmission electron microscopy (TEM) giving the relative fraction of crystals exposing the 101 face. It was found to be 3 times larger in one sample compared to the other, indicating a significantly different exposed surface. Thus, regardless of the specific size-induced property that prompts the use of nanosize materials in a particular application, surface effects can become a significant factor in the overall performance of this application. When surface-related processes are associated with the nanomaterial utilization, like in many catalysis or electrochemical applications,2,4,7 the specific surface properties of these particles become crucial. Therefore it is important to develop the ability to characterize the surface of nanoparticles in addition to the standard parameters such as size, size distribution, shape, and crystal structure, that are usually considered. Currently the ability to determine the surface structures of nanoparticles is usually limited to techniques that lack a statistical significance like electron microscopy and scanning probe techniques or to macroscopic measurements adopted from bulk systems that average the overall surface characteristics.3,4 The natural, more or less pronounced, polydispersity of inorganic nanoparticles leads to a superposition of the surface properties with the particle size distribution and is thus very difficult to address. Analytical ultracentrifugation (AUC, for a recent review see ref 9) is known as a superior technique which (8) Zaban, A.; Aruna, S. T.; Tirosh, S.; Gregg, B. A.; Mastai, Y. J. Phys. Chem. B 2000, 104, 4130.
10.1021/la0347051 CCC: $25.00 © 2003 American Chemical Society Published on Web 11/19/2003
Nanoparticle Surface Structure Analysis
Langmuir, Vol. 19, No. 26, 2003 10655
can deal very well with polydispersity as it can fractionate a sample according to size and density over the entire colloidal range with high statistical significance as every particle is detected.10,11 AUC can in certain circumstances even determine the particle size distributions of the smallest colloids, 0.02 and may lead to dispersion of rutile TiO2 even at its isoelectric point of pH ≈ 6.2 at 1 M NaCl.22 In other words, the AUC measurement that provides a distribution of sedimentation coefficients is highly sensitive to aggregation including small aggregates in small amounts. The effect is evident in the increase of the sedimentation coefficient and a change of the distribution shape. Thus comparing the sedimentation coefficient distribution of samples 1 and 2, which have a quite similar particle size, density, and shape, should correlate to the particle surface, which is highly affected by the crystal face exposed by the nanocrystals as well as so called “structural forces”, for example, the correlated local changes of the surface-bound water structure under the influence of the force fields originating from the exposed crystal faces.23 Figure 4 shows the sedimentation coefficient distributions of the two samples as a function of the solution pH (22) Yotsumotu, H.; Yoon, R. H. J. Colloid Interface Sci. 1993, 157, 426. (23) Churaev, N. V.; Derjaguin, B. V. J. Colloid Interface Sci. 1985, 103, 542.
Nanoparticle Surface Structure Analysis
Langmuir, Vol. 19, No. 26, 2003 10657
Figure 4. pH-dependent sedimentation coefficient distributions for (a) sample 2 and (b) sample 1.
starting at a low pH value. The sedimentation coefficient distributions were not corrected for the effects of diffusion broadening so that they appear broader than the real distribution. Figure 4 clearly shows that the sedimentation coefficient distributions of the two tested samples are not the same. They differ by the peak maximum, width, shape, and response to the solution pH. In the following section, we show that this behavior results from the different surfaces of the two nanocrystal samples expressed by the process of particle aggregation. With respect to other methods that provide an average charge value (like ζ-potential measurements), the AUC measurement has the advantage of yielding a high-resolution distribution. Figure 5 presents the peak maximum of the sedimentation coefficient distribution of the two nanocrystal samples at different solution pHs. Starting at pH 1, the sedimentation coefficients at peak maximum (sample 1, 1330 S; sample 2, 3120 S) differ by more than a factor of 2. Both the primary charge effect and the aggregation effect follow the same trend for a decreased electrostatic particle stabilization due to an increase in the ionic strength. As the particles appear to be similarly charged according to the ζ-potential measurements (Figure 3), the huge observed differences are a reflection of the differing particle surface structures expressed in their aggregation behavior. From Figure 5b and the comparison with the primary particle sizes (solid horizontal lines), it becomes clear that already at pH 1 the particles are aggregated, either during synthesis or washing, although to a different extent. Again, this can be attributed to a different particle surface structure of the two samples. When we increase the pH by base addition, the ionic strength of 0.1 mol/L, which is determined by the high nitric acid concentration, is not notably changed. Thus it
Figure 5. (a) Sedimentation coefficients from the peak maximum of the sedimentation coefficient distribution at different solution pHs; (b) pH-dependent particle sizes calculated on the basis of the sedimentation coefficients presented in part a with a bulk anatase density of 3.84 g/mL. The horizontal lines in part b indicate the particle diameters from electron microscopy and XRD (see above).
cannot be argued that the particle charge is screened more and more by the increase in the counterion concentration providing a lower electrostatic contribution toward electrostatic stabilization due to the decrease in the electric double layer upon the approach to the point of zero charge, pH 6.9 for sample 1 and pH 6.6 for sample 2 (ζ-potential measurements, Figure 3, above). Consequently, when purely considering the simple Derjaguin-Landau-Verwey-Overbeek (DLVO) theory (van der Waals attraction and electrostatic repulsion) for electrostatically stabilized colloids, with increasing pH and in the absence of aggregation, the s-value should be constant up to the point where the colloids undergo massive aggregation resulting in a sudden rise of the sedimentation coefficient due to chemical changes in the particle surface structure. Figure 5 shows this trend only for sample 1 in a limited pH interval revealing a pronounced difference in the surface structures of the two colloid samples. The s-value of sample 2 continuously increases with increasing pH, reaching the significant aggregation at a pH between 1.96 and 1.99. Thus, AUC clearly shows delicate differences between the two preparations, which are all related to the particle surface structure and are expressed in the following: differences in the colloidal stability; subtle charge differences, not being expressed in the ζ-potential; shape of the aggregates; particle hydration. The observed changes in the sedimentation coefficient may be attributed to differences in the colloidal stability of the two samples, which despite their similar primary
10658
Langmuir, Vol. 19, No. 26, 2003
Co¨ lfen et al.
particle size, shape, and density expose different surface structures. The resulting differences in the particle aggregation are also reflected in the volume weighed average hydrodynamic particle diameters dh and the translational diffusion coefficients D calculated from dynamic light scattering (DLS):
calculated from the assumption that the potential at the limit of the Stern layer (2 water layers, 0.5 nm) is the ζ-potential. The potential ψ in the distance x of a plane surface can be approximated by
Sample 1 (volume weighed)
which is similar to the Debye-Hu¨ckel approximation for spherical particles for low potentials (
